def test_hoeffding_anytime_tree_coverage():
# Cover memory management
stream = SEAGenerator(random_state=1, noise_percentage=0.05)
stream.prepare_for_use()
X, y = stream.next_sample(15000)
learner = HATT(max_byte_size=30,
memory_estimate_period=100,
grace_period=10,
leaf_prediction='nba')
learner.partial_fit(X, y, classes=stream.target_values)
learner.reset()
# Cover nominal attribute observer
stream = RandomTreeGenerator(tree_random_state=23,
sample_random_state=12,
n_classes=2,
n_cat_features=2,
n_categories_per_cat_feature=4,
n_num_features=1,
max_tree_depth=30,
min_leaf_depth=10,
fraction_leaves_per_level=0.45)
stream.prepare_for_use()
X, y = stream.next_sample(15000)
learner = HATT(leaf_prediction='nba',
nominal_attributes=[i for i in range(1, 9)])
learner.partial_fit(X, y, classes=stream.target_values)
def test_hoeffding_anytime_tree_nba(test_path):
stream = RandomTreeGenerator(tree_random_state=23,
sample_random_state=12,
n_classes=2,
n_cat_features=2,
n_categories_per_cat_feature=4,
n_num_features=1,
max_tree_depth=30,
min_leaf_depth=10,
fraction_leaves_per_level=0.45)
stream.prepare_for_use()
learner = HATT(nominal_attributes=[i for i in range(1, 9)])
cnt = 0
max_samples = 5000
predictions = array('i')
proba_predictions = []
wait_samples = 100
while cnt < max_samples:
X, y = stream.next_sample()
# Test every n samples
if (cnt % wait_samples == 0) and (cnt != 0):
predictions.append(learner.predict(X)[0])
proba_predictions.append(learner.predict_proba(X)[0])
learner.partial_fit(X, y)
cnt += 1
expected_predictions = array('i', [
1, 0, 1, 0, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 0, 0, 1, 0, 1, 1, 0, 1, 1, 1,
1, 0, 1, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0,
0
])
assert np.alltrue(predictions == expected_predictions)
test_file = os.path.join(test_path, 'test_hoeffding_anytime_tree.npy')
expected_proba = np.load(test_file)[:49, :]
assert np.allclose(proba_predictions, expected_proba)
expected_info = "HATT(binary_split=False, grace_period=200, leaf_prediction='nba',\n" \
" max_byte_size=33554432, memory_estimate_period=1000000,\n" \
" min_samples_reevaluate=20, nb_threshold=0,\n" \
" nominal_attributes=[1, 2, 3, 4, 5, 6, 7, 8], split_confidence=1e-07,\n" \
" split_criterion='info_gain', stop_mem_management=False,\n" \
" tie_threshold=0.05)"
assert learner.get_info() == expected_info
expected_model = 'ifAttribute1=0.0:ifAttribute3=0.0:Leaf=Class1|{0:260.0,1:287.0}' \
'ifAttribute3=1.0:Leaf=Class0|{0:163.0,1:117.0}ifAttribute1=1.0:Leaf=Class0|{0:718.0,1:495.0}'
assert (learner.get_model_description().replace("\n", " ").replace(
" ", "") == expected_model.replace(" ", ""))
assert type(learner.predict(X)) == np.ndarray
assert type(learner.predict_proba(X)) == np.ndarray
def test_hoeffding_anytime_tree_nb_gini(test_path):
stream = RandomTreeGenerator(tree_random_state=23,
sample_random_state=12,
n_classes=2,
n_cat_features=2,
n_categories_per_cat_feature=4,
n_num_features=1,
max_tree_depth=30,
min_leaf_depth=10,
fraction_leaves_per_level=0.45)
stream.prepare_for_use()
learner = HATT(nominal_attributes=[i for i in range(1, 9)],
leaf_prediction='nb',
split_criterion='gini')
cnt = 0
max_samples = 5000
predictions = array('i')
proba_predictions = []
wait_samples = 100
while cnt < max_samples:
X, y = stream.next_sample()
# Test every n samples
if (cnt % wait_samples == 0) and (cnt != 0):
predictions.append(learner.predict(X)[0])
proba_predictions.append(learner.predict_proba(X)[0])
learner.partial_fit(X, y)
cnt += 1
expected_predictions = array('i', [
1, 0, 1, 0, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 0, 0, 1, 0, 1, 1, 0, 1, 1, 1,
1, 0, 1, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0,
0
])
assert np.alltrue(predictions == expected_predictions)
expected_info = "HATT(binary_split=False, grace_period=200, leaf_prediction='nb',\n" \
" max_byte_size=33554432, memory_estimate_period=1000000,\n" \
" min_samples_reevaluate=20, nb_threshold=0,\n" \
" nominal_attributes=[1, 2, 3, 4, 5, 6, 7, 8], split_confidence=1e-07,\n" \
" split_criterion='gini', stop_mem_management=False, tie_threshold=0.05)"
assert learner.get_info() == expected_info
def basic_simulation():
model = HATT()
print("EFDT Basic Sim")
num_choices = 2
click_gen = de.default_sex_age_click_gen()
de.test_default_sex_age_click_gen()
batch_size = 1000
batch_count = 10
batches = de.get_batches(batch_size, batch_count)
choice = 0
for i, batch in enumerate(batches):
choices = [choice] * len(batch)
clicks = [click_gen.get_click(visit, choice) for visit,choice in zip(batch, choices)]
print_clicks(de.types, i, clicks, batch, choices)
tim = time.time()
model.partial_fit(batch, clicks)
print("fit time: ", time.time() - tim)
print("--Model: ", choice)
print_predictions(model)
def test_hoeffding_anytime_tree(test_path):
stream = RandomTreeGenerator(tree_random_state=23,
sample_random_state=12,
n_classes=2,
n_cat_features=2,
n_categories_per_cat_feature=4,
n_num_features=1,
max_tree_depth=30,
min_leaf_depth=10,
fraction_leaves_per_level=0.45)
stream.prepare_for_use()
learner = HATT(nominal_attributes=[i for i in range(1, 9)])
cnt = 0
max_samples = 15000
predictions = array('i')
proba_predictions = []
wait_samples = 100
while cnt < max_samples:
X, y = stream.next_sample()
# Test every n samples
if (cnt % wait_samples == 0) and (cnt != 0):
predictions.append(learner.predict(X)[0])
proba_predictions.append(learner.predict_proba(X)[0])
learner.partial_fit(X, y)
cnt += 1
expected_predictions = array('i', [
1, 0, 1, 0, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 0, 0, 1, 0, 1, 1, 0, 1, 1, 1,
1, 0, 1, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0,
0, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 1, 1, 0, 1, 0, 0,
1, 0, 1, 1, 1, 1, 0, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0,
1, 1, 0, 1, 1
])
test_file = os.path.join(test_path, 'test_hoeffding_anytime_tree.npy')
data_prob = np.load(test_file)
assert np.alltrue(predictions == expected_predictions)
assert np.alltrue(proba_predictions == data_prob)
expected_info = 'HATT: max_byte_size: 33554432 - memory_estimate_period: 1000000 - grace_period: 200 - ' \
'min_samples_reevaluate: 20 - split_criterion: info_gain - split_confidence: 1e-07 - ' \
'tie_threshold: 0.05 - binary_split: False - stop_mem_management: False - leaf_prediction: ' \
'nba - nb_threshold: 0 - nominal_attributes: [1, 2, 3, 4, 5, 6, 7, 8] - '
assert learner.get_info() == expected_info
expected_model = 'if Attribute 1 = 0: if Attribute 3 = 0: Leaf = Class 1 | {0: 896.0, 1: 947.0} ' \
'if Attribute 3 = 1: Leaf = Class 0 | {0: 500.0, 1: 388.0} if Attribute 1 = 1: ' \
'if Attribute 5 = 0: Leaf = Class 0 | {0: 404.0, 1: 259.0} if Attribute 5 = 1: ' \
'Leaf = Class 0 | {0: 166.0, 1: 82.0}'
assert (learner.get_model_description().replace("\n", " ").replace(
" ", "") == expected_model.replace(" ", ""))
assert type(learner.predict(X)) == np.ndarray
assert type(learner.predict_proba(X)) == np.ndarray
class Tree():
def check_input(self, kwargs):
if all(k in kwargs for k in ['tk', 'provider']):
return isinstance(kwargs['tk'], tkinter.Tk) and isinstance(
kwargs['provider'], DataProvider)
def __init__(self, **kwargs):
if not self.check_input(kwargs):
raise ValueError(
'Tree expects two arguments: \'tk\', a Tkinter object, and \'provider\' a DataProvider object'
)
# save arguments
self.tk = kwargs['tk']
self.provider = kwargs['provider']
# subscribe to algorithm and listeners lists
self.provider.subscribe_to_listening_list(self)
self.provider.subscribe_to_algorithm_list(self)
# list to save all tree states (store root nodes)
self.history = []
self.current_timestamp = -1
# get the actual ExtremlyFastDecisionTreeClassifier here
self.tree = TreeClass()
# master frame
self.frame = tkinter.Frame(self.tk,
highlightthickness=5,
highlightbackground='black',
bd=0)
self.frame.pack(side=tkinter.TOP, fill=tkinter.BOTH)
def _save_tree_state(self):
self.history.append(Node(self.tree._tree_root))
def notify(self):
# destroy previous widgets
if (self.current_timestamp != -1):
self.history[self.current_timestamp].destroy()
# update 'self.current_timestamp' and build widgets for the new timestamp
self.current_timestamp = self.provider.get_timestamp()
self.history[self.current_timestamp].build_widget(self.frame)
def train(self):
# just save the new tree state to 'self.history', 'notify' will take care of the widget
# train tree using current object
obj = self.provider.get_current_object()
self.tree.partial_fit(obj['x'], obj['y'])
# save new tree
self._save_tree_state()